Quantitative assessment of fish larvae community composition in spawning areas using metabarcoding of bulk samples

Abstract

Accurate assessment of larval community composition in spawning areas is essential for fisheries management and conservation but is often hampered by the cryptic nature of many larvae, which renders them difficult to identify morphologically. Metabarcoding is a rapid and cost‐effective method to monitor early life stages for management and environmental impact assessment purposes but its quantitative capability is under discussion. We compared metabarcoding with traditional morphological identification to evaluate taxonomic precision and reliability of abundance estimates, using 332 fish larvae from multinet hauls (0–50 m depth) collected at 14 offshore sampling sites in the Irish and Celtic seas. To improve quantification accuracy (relative abundance estimates), the amount of tissue for each specimen was standardized and mitochondrial primers (12S gene) with conserved binding sites were used. Relative family abundance estimated from metabarcoding reads and morphological assessment were positively correlated, as well as taxon richness (RS = 0.81, P = 0.007) and diversity (RS = 0.90, P = 0.002). Spatial patterns of community composition did not differ significantly between metabarcoding and morphological assessments. Our results show that DNA metabarcoding of bulk tissue samples can be used to monitor changes in fish larvae abundance and community composition. This represents a feasible, efficient, and faster alternative to morphological methods that can be applied to terrestrial and aquatic habitats.

Full text

Quantitative assessment of fish larvae community composition in
spawning areas using metabarcoding of bulk samples
FRANCES C. RATCLIFFE,TAMSYN M. UREN WEBSTER,DEIENE RODRIGUEZ-BARRETO,RICHARD O’RORKE,
CARLOS GARCIA DE LEANIZ,AND SOFIA CONSUEGRA
1
Department of Biosciences, College of Science, Swansea University, Swansea SA2 8PP UK
Citation: Ratcliffe, F. C., Uren Webster, T. M., Rodriguez-Barreto, D., O'Rorke, R., Garcia de Leaniz, C.,
and Consuegra, S. 2021. Quantitative assessment of fish larvae community composition in spawning areas
using metabarcoding of bulk samples. Ecological Applications 31(3):e02284. 10.1002/eap.2284
Abstract. Accurate assessment of larval community composition in spawning areas is
essential for fisheries management and conservation but is often hampered by the cryptic nat-
ure of many larvae, which renders them difficult to identify morphologically. Metabarcoding is
a rapid and cost-effective method to monitor early life stages for management and environ-
mental impact assessment purposes but its quantitative capability is under discussion. We com-
pared metabarcoding with traditional morphological identification to evaluate taxonomic
precision and reliability of abundance estimates, using 332 fish larvae from multinet hauls
(0–50 m depth) collected at 14 offshore sampling sites in the Irish and Celtic seas. To improve
quantification accuracy (relative abundance estimates), the amount of tissue for each specimen
was standardized and mitochondrial primers (12S gene) with conserved binding sites were
used. Relative family abundance estimated from metabarcoding reads and morphological
assessment were positively correlated, as well as taxon richness (R
S
=0.81, P=0.007) and
diversity (R
S
=0.90, P=0.002). Spatial patterns of community composition did not differ sig-
nificantly between metabarcoding and morphological assessments. Our results show that DNA
metabarcoding of bulk tissue samples can be used to monitor changes in fish larvae abundance
and community composition. This represents a feasible, efficient, and faster alternative to mor-
phological methods that can be applied to terrestrial and aquatic habitats.
Key words: 12S; bulk samples; Celtic sea; fish larvae; Irish sea; metabarcoding; quantification.
INTRODUCTION
Assessing larval community composition is needed to
provide accurate information about spawning areas for
fisheries management and conservation, but the location
and dispersal of larval stages are largely unknown
aspects of many fish life cycles (Legrand et al. 2019).
Early life stages of organisms are particularly sensitive
to abiotic stressors (Radchuk et al. 2013) and, for fish,
understanding the quantitative relationship between
environmental quality and population dynamics remains
challenging (Rose 2000). Thus, larval monitoring pro-
vides critical information about population changes over
time (Asch 2015) to inform conservation and policy
(Ellis et al. 2012, Borja et al. 2017), but its application is
often hampered by the cryptic morphology of early life-
stage organisms (Brechon et al. 2013, Sigut et al. 2017,
Kimmerling et al. 2018).
Traditional fish larvae monitoring involves identifying
each individual using a light microscope, counting
myotomes, assessing pigmentation patterns and jaw
morphology (Russell 1976). Yet, identification keys are
incomplete for many parts of the world (Becker et al.
2015) and, where descriptions are available, morphologi-
cal assessment is time consuming and requires specialist
training (Brechon et al. 2013). Morphological taxonomy
also relies on the identifying features remaining intact
for species level assignment (Russell 1976), but damage
is common during sampling (e.g., when using continuous
plankton recorders), leading to misidentification and
loss of valuable information (Richardson et al. 2006).
In cases where morphological identification is unfea-
sible, DNA sequencing technologies may be used to
identify organisms, as long as their sequences are in the
databases (Taberlet et al. 2012). The development of
high-throughput sequencing technology allows ampli-
con-based sequencing (metabarcoding) of multiple
individuals of various species concurrently (i.e., bulk
samples), providing a relatively quick method of pro-
cessing many samples to obtain taxonomical informa-
tion (Taberlet et al. 2012) and estimate biodiversity
(Dopheide et al. 2019). However, obtaining accurate
absolute abundance (number of individuals) estimates
through relative read abundance (RRA) from amplicon
sequence data has remained challenging (Deagle et al.
Manuscript received 5 May 2020; revised 10 August 2020;
accepted 5 October 2020. Corresponding Editor:
´
Eva E. Pla-
ganyi.
1
Corresponding Author. E-mail: s.consuegra@swansea.ac.uk
Article e02284; page 1
Ecological Applications, 31(3), 2021, e02284
©2021 The Authors. Ecological Applications published by Wiley Periodicals LLC on behalf of Ecological Society of America.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and
distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

2019, Lamb et al. 2019). This is because biases in RRA
estimations can be introduced at different stages of the
metabarcoding protocol, for example, cell and DNA
quantity, mitochondrial copy number, extraction suc-
cess and PCR amplification rates can vary between tis-
sue typ e and spec ies (Lamb et al. 2019, Pi ˜
nol et al.
2019), leading to inaccurate estimates. Another source
of bias can arise from unequal body size of individuals
pooled within a bulk sample, which can be mitigated by
size fractioning of organisms prior to extraction,
increasing the reliability of RRA estimates (Elbrecht
et al. 2017). The choice of primers and target region
may introduce further bias (Deagle et al. 2014). These
biases have led to designing costly and bioinformati-
cally challenging metagenomic approaches (Tang et al.
2015, Kimmerling et al. 2018) or to the use of multiple
loci (Richardson et al. 2015) to identify particular spe-
cies and estimate their abundance.
Improving the reliability of abundance estimates is
thus needed to make metabarcoding more useful for bio-
diversity monitoring, calculation of metrics such as
diversity indices, as well as detection of natural shifts in
multispecies community composition (Bohmann et al.
2014). Different approaches have been proposed to
improve abundance estimates based on RRA, while still
using a cost-effective, single-marker, PCR approach
(Thomas et al. 2016, Elbrecht et al. 2017). For example,
using primers with widely conserved priming sites may
reduce taxa specific biases (Krehenwinkel et al. 2017),
although taxonomic resolution can be reduced due to
highly similar sequences within a family (Thomsen et al.
2016).
Here, using a single mitochondrial marker (12S ribo-
somal RNA, considered highly specific in fish), we have
refined the reliability of DNA metabarcoding abundance
estimates by standardizing input material and choosing
conserved primer binding sites. Using bulk fish larvae
samples from the Irish and Celtic Seas, we compared the
sensitivity and accuracy of this approach with tradi-
tional morphological identification, to assess whether
metabarcoding can be a feasible and rapid alternative to
traditional assessment for estimating fish larvae richness,
diversity, and community composition metrics.
MATERIALS AND METHODS
Field sampling
Sampling was carried out onboard the RV Celtic Voy-
ager between 17 and 26 May 2018. Fish larvae
(3–30 mm) from 14 hauls (one per site) were sampled
using a MultiNet plankton sampler (Hydro-Bios, Kiel,
Germany). Sites 1–8 and 12 were sampled with one obli-
que haul to 50 m depth per site, filtering a mean volume
of 215 55 m
3
of water. Hauls 9–14 (with the exception
of haul 12) consisted of two vertical hauls from the sur-
face to 50 m, filtering a mean volume of 38 6m
3
,
which were pooled for each site. Fish larvae from each
haul were separated from other zooplankton species and
preserved in RNAlater (Qiagen, Hilden, Germany) at
room temperature for 24 h, then refrigerated at 4°C until
morphological identification.
Morphological identification
Fish larvae ranged from 2 to 30 mm total length. For
morphological identification, all larvae were first sepa-
rated into major groupings based on body shape follow-
ing the classification by Russell (1976) and subsequently
assigned to family level. Assignment to genus and spe-
cies where possible, was then carried out. Assignments
were checked against the species descriptions first in
Russell (1976), and, where possible, double checked
against the description by Rodriguez et al. (2017). For
taxa that could not be confidently morphologically iden-
tified, DNA was extracted from one or more representa-
tive individuals (34 individuals of 16 taxa across the
survey, Appendix S1: Table S1) using the Qiagen
DNeasy Blood and Tissue kit (Qiagen GmbH) following
the manufacturer’s instructions. Extracted DNA was
then amplified using the 12S V5 primers (Riaz et al.
2011a), cleaned using a sodium acetate/EtOH solution,
resuspended in 10 µL HiDi Formamide (Applied Biosys-
tems, Foster City, CA, USA) and analysed using Sanger
Sequencing on an ABI 3730 DNA Analyzer (Applied
Biosystems). Resulting sequences were aligned in Bio-
Edit v 7.2.5 (Hall et al. 2011). When 12S barcoding did
not resolve taxonomic identification to species level, due
to database limitations or synonymous sequences, the
barcoding region of ~650 bp of the CO1 gene (F1, R1;
Ward et al. 2005) was used to update taxonomic assign-
ment to the lowest possible taxonomic level, resulting in
six additional 12S reference sequences not present in the
NCBI nucleotide database (Appendix S1: Table S1;
Genbank accession numbers: MN539950, MN539961,
MN539952, MN539964, MN539965, MN539966). Tax-
onomy of Sanger sequenced individuals was assigned to
the lowest possible level using the MegaBLAST algo-
rithm (Morgulis et al. 2008) against the National Center
for Biotechnology Information (NCBI) GenBank
nucleotide database (accessed November 2018) and dou-
ble checked against the BOLD database (https://
www.boldsystems.org/). To estimate accuracy and
repeatability of taxonomic assignments, a group of 15
specimens were also sent to an experienced taxonomist
and verified by CO1 barcoding (morphological taxo-
nomic assignment concordance test).
DNA extraction
After taxonomic identification, bulk tissue samples
from all larvae of each haul were prepared for DNA
extraction as follows: 2–8 mg of tissue were cut from the
area anterior to the tail of each juvenile fish (for individ-
uals <5 mg, the entire larva was used, n=88) and
placed in a Falcon tube on ice. Buffer ATL and
Article e02284; page 2 FRANCES C. RATCLIFFE ET AL. Ecological Applications
Vol. 31, No. 3

proteinase K (Qiagen DNeasy Blood and Tissue kit)
were then added to the pooled tissue sample in a ratio of
180 µL of ATL and 20 µL proteinase K for 15 mg of tis-
sue. Each falcon tube (representing one haul) was vor-
texed thoroughly and incubated overnight to digest at
56°C, shaking at 65 rpm. Samples were visually
inspected for tissue remnants, vortexed, and re-incu-
bated until all tissue dissolved. Digestions from each
haul were then vortexed for 45 s to ensure thorough mix-
ing of digested products and divided in three sub-sam-
ples of 200 µL that were extracted using the Qiagen
DNeasy Blood and Tissue kit, following the manufac-
turer’s instructions. Extraction blanks were carried
through each step of the process.
Library preparation and sequencing
A 106-bp fragment of the 12S mitochondrial was
amplified with the 12S V5 primers (Riaz et al. 2011b)
using Phusion High-Fidelity DNA Polymerase (Thermo
Fisher Scientific, Loughborough, UK), with an anneal-
ing temperature of 52°C, in three extraction replicates
per haul. Libraries were prepared using a two-step PCR
approach, based on the Illumina 16S Metagenomic
Sequencing Library preparation guidelines (Illumina,
San Diego, California, USA), with following adapta-
tions: in the first PCR step, each extraction replicate was
amplified in triplicate in order to increase detection of
rare species (Alberdi et al. 2018). Subsequently, 10 µL
from each triplicate were pooled prior to first cleanup.
Cleanups were performed using Agencourt AMPure XP
beads (Beckman Coulter, Brea, California, USA), using
a 1.2 ×volume of beads to PCR product. Amplicons
were indexed using Nextera XT Index Kit v2 Set C (Illu-
mina), and DNA concentration of each reaction was
quantified via Qubit dsDNA HS Assay (Invitrogen,
Carlsbad, California, USA) and pooled in equal molar
concentrations. PCR and extraction blanks (using
molecular grade water instead of template) were sub-
jected to all steps of the library preparation process. In
addition, a sequencing/tag jumping blank, where no
sample was added prior to sequencing, was used. Pair-
end sequencing was carried out at Swansea University
using an Illumina MiSeq platform (Illumina)
(2 ×300 bp reads), including 5% PhiX.
Bioinformatics: sequence processing
De-multiplexed samples containing raw pair-end
sequences were processed using Qiime2 (version 2019.1;
Bolyen et al. 2019). Initially, raw sequences were quality
checked using interactive quality plots, in order to
obtain values for sequence trimming and truncation.
De-noising was carried out using DADA2 (Callahan
et al. 2017) where the first 10 bp of each sequence were
trimmed to remove adaptors and all sequences truncated
to 100 bp in length based on quality scores. Default
DADA2 settings within Qiime2 were used to detect and,
where possible, correct sequencing errors and filter out
phiX reads and chimeric sequences, join pair-end reads,
and de-replicated sequences. The amplicon sequence
variant (ASV) approach was chosen because it provides
a higher resolution than a traditional OTU approach,
enabling detection of single nucleotide differences (Cal-
lahan et al. 2017). After de-noising, the ASV and
BIOME tables were exported for taxonomic assignment.
Database construction and taxonomic assignment
A custom database was constructed using in silico
PCR against the NCBI database (downloaded February
2019): 12S V5 primers were allowed to have three base
mismatches in silico (search_PCR command; Edgar
2010) and a corresponding taxonomy file was con-
structed using the obiannotate tool (OBITools; Boyer
et al. 2016). All sequences were trimmed to the target
region. A list of all marine fish species encountered in
the British Isles, including nonnative fish (366 species;
Fish Base: accessed 31 March 2019) was then used to fil-
ter the main database to fish species present in the study
region, of which 207 were available. The six 12S Sanger
sequences (generated with the 12S V5 primers) missing
from NCBI database and verified using CO1 barcoding
from this study were added to the database (Appendix
S1: Table S1), which also included marine mammals,
bacteria, and other contaminants (such as Homo sapi-
ens) that might be amplified by the primers.
Initially, ASVs were classified using the KNN method
in Mothur (Schloss et al. 2009) using the parameter
numwanted =1 (Findley et al. 2013), against the custom
database. Because this parameter may lead to false posi-
tive assignments, KNN assignments were then verified
using NCBI megaBLAST, with max-target sequences =
10. The top 10 assignments were screened for UK spe-
cies (Fish Base) on a case by case basis. Where the per-
centage of UK species match fell below 98%, or where
multiple UK species matched above a 98% match,
MEGAN (6.15.1) was were used to assign species to the
lowest common ancestor (Huson et al. 2007). ASVs for
which there were no vertebrate matches were discarded
from downstream analysis.
Tag jumping/cross-contamination (Schnell et al. 2015)
was removed on the following basis: a taxon was
removed from a haul if it had fewer than 115 reads (max-
imum reads for a single species in tag jumping control
sample) or did not appear in all three replicates.
For spatial analysis, numbers of individuals of each
taxon in a haul were estimated from the proportion of
reads in the corresponding sample, as follows:
Ai¼NPi(1)
where A
i
is the abundance (number of individuals) of the
taxon of interest (i) in a given haul, Nis total number of
individuals in the haul, and P
i
is the proportion of that
taxon in the haul amplicon pool.
April 2021 QUANTITATIVE FISH LARVAE METABARCODING Article e02284; page 3

Statistical analysis
The accuracy of estimates of RRA and diversity indices
derived from metabarcoding was assessed against results
from morphological taxonomy using Spearman’srank
correlation analysis performed in R version 3.5.2 (R Core
Team 2020). Diversity indices (Shannon Weiner Index,
Simpson’s Diversity) and richness were estimated based
on RRA and morphological relative abundances using
the Vegan package (R version 3.5.2) for both lowest possi-
ble taxonomic and family level taxon identifications. For
spatial analysis, the survey area was divided into three
locations along a temperature gradient: Loc 1 (above the
Celtic/Irish sea front, 9°–10.99°C), Loc 2 (channel spawn-
ing grounds, 11°–12.99°C), and Loc 3 (western Celtic Sea,
13–14°C; Fig. 1). The number of individuals (assessed
morphologically) and estimated from reads (Eq. 1), of a
given taxon (mean of the three technical replicates per
site) were divided by the volume of water filtered in the
corresponding haul (Canfield and Jones 1996) to obtain
catch per unit filtered (CPUF) or estimated number of
individuals from reads per unit filtered (RPUF) values,
respectively. This analysis was carried out at both lowest
possible taxonomic level and family level. All 14 hauls
surveyed were included in this analysis, where only one
individual was present in a haul this was divided by the
volume of water filtered and included in the both the
CPUF and RPUF data sets. The family Ammodytidae
was excluded from this analysis, because not all individu-
als were retained in haul 4. CPUF and RPUF values were
square-root transformed, and composition similarity cal-
culated by hierarchical clustering using a Bray-Curtis
resemblance matrix. Subsequently, pairwise analysis of
similarities (ANOSIM) was used to test whether there
was a significant difference in community composition
between locations (Clarke 1993), using both the CPUF
and RPUF methods. Where significant differences were
detected, SIMPER analysis (Clarke 1993) was used to
ascertain which taxa accounted for the differences
observed. Diversity indices calculations and multivariate
spatial analyses were performed using Primer-v7 (Clarke
and Gorley 2015).
RESULTS
Morphological assessment
A total of 332 fish larvae were caught in 11 of the 14
hauls in the survey. No larvae were encountered in hauls
10, 11, and 14 and only one in hauls 1 and 6, therefore 9
of the 14 hauls were used in metabarcoding. The maxi-
mum number of individuals per haul was 63 (haul 2)
(Appendix S1: Table S2). Morphological identification
assigned 324 (98%) of individuals to family level. It was
not possible to assign the families of the remaining eight
larvae, due to damaged identifying features. Of those
specimens assigned to family level, 255 (77%) were
assigned to a genus and 100 (30%) to a species. Sanger
sequencing to check morphological assignment compris-
ing of 34 individuals across nine hauls), contained 15
taxa (Appendix S1: Table S3). In the morphological tax-
onomic assignment test of the 15 individuals identified
by two independent observers and subsequently checked
by Sanger sequencing, 100% and 93% were correctly
assigned to family level by the first and second observer
respectively, 86.3% and 53.3% to genus and 40% to spe-
cies level in both cases.
Based on morphology alone, before verification with
CO1/12S barcoding, taxa within Ammodytidae and Clu-
peidae could not be assigned further than family level.
Most clupeids did not amplify with the CO1 primers and
those that did were assigned to S. sprattus; for Calliony-
mus there was no C. reticulatus sequence to compare with.
Incorrect morphological assignments occurred in the cases
of Micromesistius poutassou (Sanger seq: M. merlangius),
Aphia minuta (Sanger seq: C. harengus/S. sprattus)and
Mugilidae (Sanger seq: L. bergylta and C. mustela)
(Fig. 2, Appendix S1: Table S3). In addition, Sardina pil-
chardus,Labrus mixtus/bergylta,Molva molva, and a taxon
belonging to the Gobiidae family were only detected using
sequencing. In contrast, M. merlangus and Pollachius
virens/pollachius, were identifiable through morphology,
but not resolved to species level by 12S metabarcoding
due to lack of variability of the 12S fragment. C. harengus
and S. sprattus could not be separated by morphology or
12S metabarcoding.
Metabarcoding assessment
A total of 3,398,391 raw 300 bp pair-end reads were
generated for this study. After Qiime2 DADA de-noising,
a total of 2,675,140 reads remained for downstream analy-
sis. Once the taxonomic assignment was complete, reads
likely present due to tag jumping from concurrent sample
sequencing (Solidae 274 reads, Scomber scombrus,10
reads, Salmo salar, 3 reads), and human reads (2,338) were
removed from downstream analysis. A total of 49 fish
ASVs remained for downstream analysis. Samples con-
tained a mean of 93,223 reads (standard deviation =
31,866) post-filtering, the tag jumping blank contained
146 reads, the PCR blank 64 reads, and extraction blanks
116 and 71 reads, respectively. Tag jumping read removal
resulted in 0.046% of reads being excluded from down-
stream analysis across the samples in the study. Post-filter-
ing, taxa distribution was concordant among the three
haul replicates in all nine hauls (Fig. 3).
Comparison of abundance estimates by morphology and
metabarcoding
The relative abundance (percentage) of individuals
identified morphologically in a sample and the corre-
sponding RRAs were positively correlated for all families
assessed (Spearman’s rank: Ammodytidae R
S
=0.93,
P<0.001; Callionymidae R
S
=0.99, P<0.001; Clupei-
dae R
S
=0.97, P<0.001; Gadidae R
S
=0.95,
Article e02284; page 4 FRANCES C. RATCLIFFE ET AL. Ecological Applications
Vol. 31, No. 3

P<0.001; Pleuronectidae R
S
=0.68, P=0.05; Triglidae
R
S
=0.88, P=0.002; Appendix S1: Fig. S1). In addi-
tion, no difference in diversity and taxon richness were
detected between the relative abundance of morphological
assignments and RRA assignments at either lowest possi-
ble taxonomic level or family level, across hauls (lowest
possible taxonomic level; Spearman’s rank, richness,
R
S
=0.84, P=0.005; Shannon index, R
S
=0.90,
P=0.002; Simpson’sdiversity,R
S
=0.90, P=0.002;
family level, Spearman’s rank, richness, R
S
=0.93,
P<0.001; Shannon index, R
S
0.91, P=0.001; Simp-
son’s diversity, R
S
=0.80, P=0.01; Fig. 4).
Spatial distribution of larvae assessed by both methods
Assessment of patterns in community composition
yielded comparable results from morphological and
metabarcoding assessment at both lowest possible taxo-
nomic level and family level. Catch per unit filtered
(CPUF) and back-estimated reads per unit filtered
(RPUF), were no different between locations 1 and 2
and 1 and 3, although locations 2 and 3 differed in com-
position (lowest possible taxonomic level ANOSIM,
CPUF R=0.233, P=0.039; RPUF R=0.209,
P=0.045; family-level ANOSIM, CPUF R=0.22,
P=0.041; RPUF R=0.205, P=0.048; Table 1). SIM-
PER analysis (percent cumulative dissimilarity contribu-
tion) attributed 48.39% (CPUF) and 42.82% (RPUF) of
the difference in composition between locations 2 and 3
to three taxa: C. harengus/S. sprattus (CPUF 21.42%,
RPUF 15.74%), Triglidae (CPUF 14.37%, RPUF
14.43%), and Callionymus (CPUF 12.61%, RPUF
12.65%) (Appendix S1: Table S4, Fig. S2). The greatest
difference observed in dissimilarity contributions for the
remaining, less abundant taxa was 2.7% (C.mustela).
This pattern was repeated at the family level (Appendix
S1: Table S4, Fig. S3).
DISCUSSION
Here, we demonstrate that metabarcoding is a reliable
and practical alternative to traditional morphological
assessment. We show that RRA estimates can be
achieved by standardizing the amount of tissue analysed
per specimen and choosing primers with conserved bind-
ing sites. These estimates can then be used to successfully
calculate diversity and community composition metrics
needed to monitor changes over time. Although more
costly in terms of consumables and sequencing, metabar-
coding involved considerably less time than
FIG. 1. Multinet haul locations in the Irish and Celtic seas. Locations for spatial analysis, based on SST, are indicated as Loc 1
(above the Celtic/Irish sea front: 9°–10.99°C), Loc 2 (channel spawning grounds: 11°–12.99°C), and Loc 3 (western Celtic Sea:
13°–14°C).
April 2021 QUANTITATIVE FISH LARVAE METABARCODING Article e02284; page 5

morphological identification, particularly for those cases
that required additional barcoding to refine the morpho-
logical identification. We could use all the individuals
collected for metabarcoding, irrespective of their preser-
vation state, while the presence of damaged or poorly
preserved specimens made difficult or even impossible
their morphological identification, particularly at the
species level.
There is considerable debate over whether amplicon
sequencing can deliver reliable quantitative data (Deagle
et al. 2019). The reliability of abundance estimates from
metabarcoding varies considerably between studies, with
some showing only a weak correlation between RRA
and abundance (Lamb et al. 2019, Pi˜
nol et al. 2019).
Still, information from RRA tends to be more informa-
tive than presence/absence assessments (Deagle et al.
2019). In contrast, metagenomic approaches that do not
require PCR amplification can successfully estimate
abundance (Kimmerling et al. 2018), although the costs
and bioinformatic complexity of this approach may be
prohibitive in many contexts (Porter and Hajibabaei
2018). As shown here, amplicon sequencing can be used
to estimate abundance and we suggest that further
refinements in metabarcoding abundance estimates will
enable wider application of amplicon sequencing.
We have shown that the use of approximately equal
weights of tissue per individual can improve RRA and
diversity estimates. Approaches based on photographi-
cally assessing the surface area of taxa and modeling
biomass might also eliminate the need for weighing tis-
sue (Kimmerling et al. 2018), although not necessarily
reducing time and costs. There are, however, several fac-
tors that can bias RRA estimates. For instance, mito-
chondrial copy number can vary, not only between
different species (Pi˜
nol et al. 2015), but between different
tissue types (Wiesner et al. 1992). We mainly used the
region anterior to the tail of each larva to account for
one of these biases (tissue type) as much as possible but
interspecific biases remain a challenge to the quantita-
tive capabilities of metabarcoding techniques (Deagle
et al. 2019). Where they are consistent for a given taxon
across all samples within a study, correction factors may
unknown
damaged
f__Triglidae
f__Solidae
g__Lepidorhombus
s__Microstomus_kitt
s__Limanda_limanda
s__Glyptocephalus_cynoglossus
f__Pleuronectidae
f__Mugilidae
s__Merluccius_merluccius
s__Molva_molva
s__Labrus_mixtus
s__Labrus_bergylta
s__Aphia_minuta
s__Buenia_jeffreysii
f__Gobiidae
s__Ciliata_mustela
s__Trisopterus_minutus
s__Trisopterus_esmarkii
g__Trisopterus
s__Pollachius_virens
s__Pollachius_pollachius
x__P.pollachius/virens_M. merlangus
s__Micromesistius_poutassou
s__Merlangius_merlangus
s__Sardina_pilchardus
x__C. harengus_ S.sprattus
g__Callionymus
g__Gymnammodytes
s__Ammodytes_marinus
f__Ammodytidae
value
75
50
25
25
50
75
100
125
value
5 x 105
4 x 105
3 x 105
2 x 105
1 x 105
0
Reads
Reads
Morphology
125
100
75
50
25
25
50
75
100
125
Morphology
ABC
Morphology
alone
Morphology +
Sanger seq
Metabarcoding
(reads)
5 x 105
4 x 105
3 x 105
2 x 105
1 x 105
0
FIG. 2. Overview of larval detections during the survey. (A) Taxonomic assignments using morphology alone (presence/ab-
sence). (B) Morphological taxonomic assignments updated with Sanger sequencing, diamonds represent total number of larvae of a
taxa observed during the survey. (C) Metabarcoding taxonomic assignments, circles represent total number of reads obtained for
each taxa, post-filtering.
Article e02284; page 6 FRANCES C. RATCLIFFE ET AL. Ecological Applications
Vol. 31, No. 3

FIG. 3. Comparison of relative read abundances (three replicates per haul, a, b, c samples) and morphological taxonomic
assignments, corrected by Sanger sequencing (one per haul, morph samples). Taxa beginning with “f__”indicates family-level
assignment, “s__”indicates species-level, and “x__”indicates two to three possible species assignments. Morphological assignments
of P. pollachius/viens,M. merlangus were grouped and morphologically assigned Glyptocephaus cycnoglossus has been reassigned to
Pleuronectidae to match metabarcoding assignments to aid visual interpretation of abundances.
FIG. 4. Consistency of diversity metrics between relative abundances of morphological assignments and relative read abundance
assignments, post bioinformatic filtering (mean of three technical replicates per site, for nine sites in the study where >1larvaewas
found), for (a) species richness, (b) Shannon Wiener diversity index, and (c) Simpson’s diversity (1 −λ). R
S
, Spearman’srankρvalues.
April 2021 QUANTITATIVE FISH LARVAE METABARCODING Article e02284; page 7

be applied (Thomas et al. 2016, Krehenwinkel et al.
2017). While we found differences in relative abundance
between morphological assessment and metabarcoding,
they did not impact the calculation of diversity and com-
munity metrics. For example, estimates of the number of
individuals of Ammodytidae differed by seven individu-
als (SD 9.32) compared to those assessed morphologi-
cally but this was not sufficient to influence community
composition. However, for applications where exact
numbers of individuals are needed (e.g., census of partic-
ular species) these differences may require consideration.
While there is no perfect marker for all studies (Deagle
et al. 2014), we have shown here the benefits of using pri-
mers with well conserved binding sites, particularly for
RRA estimates. Whereas the CO1 marker has extensive
sequence databases as well as a strong capability to dis-
criminate between species, it also carries an increased
risk of amplification bias due to the lack of conserved
binding sites across a broad range of taxa (Deagle et al.
2014). This can result in false negatives where taxa
known to be present in a sample do not amplify (Collins
et al. 2019, Nobile et al. 2019). Using more conserved
priming sites, such as the 12S marker, may reduce taxa
specific biases (Krehenwinkel et al. 2017), although it
has been argued that taxonomic resolution may be
reduced due to lack of sequence variability within fami-
lies (Thomsen et al. 2016), and the completeness of refer-
ence databases also influences the resolution to species
level (Miya et al. 2015). Here, using 12S primers, 40% of
the taxa identified with metabarcoding could be assigned
to species level, with the rest being assigned to family or
genus level. In comparison to morphological identifica-
tion without the assistance of CO1 Sanger sequencing,
12S metabarcoding achieved higher taxonomic resolu-
tion and more accurate identifications to family level.
Morphologically assessed groupings supported by bar-
coding with Sanger sequencing achieved a similar level
of assignment at the family level to metabarcoding
across the study. Yet, while short reads can struggle to
resolve some families to species level (Thomsen et al.
2016), hindering species level data interpretation, we
found that the use of metabarcoding improved taxo-
nomic assignment overall. Morphology only performed
better than sequencing in the case of Glyptocephalus
cynoglossus, due to distinct morphological characteris-
tics, and in a few cases due to lack of information or
sequence variation at the 12S region. In general, synony-
mous sequences at the target region resulted in just two
(e.g., C. harengus/S. sprattus) or three species (M. mer-
langus/P. pollachius/P. virens) not being distinguished
from each other. For studies requiring species-level iden-
tification, taxa affected by lack of marker sequence
information, or variability, a targeted qPCR approach
(Brechon et al. 2013), similar to those carried out to
detect particular species using eDNA (Robinson et al.
2018) or a family-specific, multi-primer approach (Riaz
et al. 2011b) could be easily used to refine metabarcod-
ing assignments. Combining different markers, as we
have done here with the 12S metabarcoding and the
CO1 barcoding, can be used to refine the databases by
adding novel sequences and by separating species that
cannot be identified based solely on small fragments.
The completeness of the database used as a reference is
critical for the accuracy of the taxonomic assignments
and, while databases are continuously increasing in size
for the most common metabarcoding makers, given the
large diversity of fish and the increasingly lower cost of
sequencing, focusing on full mitochondrial genomes
may have wider relevance (Collins et al. 2019). We found
that in some cases metabarcoding could not resolve
identifications to species level, however, for some appli-
cations, genus level analysis provides similar diversity
and community composition information than species
level and would be appropriate, for example to detect
responses to environmental change (Hernandez et al.
2013). In some other cases, family-level analysis has been
deemed sufficient to detect broadscale changes, e.g.,
after major environmental disturbance (Hernandez et al.
2013). Therefore, dependent on hypothesis, a single 12S
analyses or an additional qPCR can be performed.
Spatial patterns detected in community composition
remained the same, independent of whether they were
assessed using morphological (CPUF) or metabarcoding
(RPUF). The small differences in abundance of rare taxa
(Appendix S1: Table S4), were mainly the result of miss-
identification of C. mustela during morphological iden-
tification, indicating that metabarcoding of bulk samples
may be used as a viable alternative to morphological
identification of samples, particularly when the latter
proves difficult.
All taxa detected in the survey were known to spawn
in the survey area (Acevedo et al. 2002, Ellis et al. 2012)
and for the family Ammodytidae, difficult to survey and
data limited due to its cryptic morphology (Ellis et al.
2012), metabarcoding identified A. marinus and a spe-
cies of the genus Gymnammodytes, further illustrating its
potential for detecting cryptic species.
TABLE 1. ANOSIM matrix, showing Rvalues of pairwise
comparisons of community composition between three
locations in the Irish/Celtic seas, using morphological
taxonomic assignments and abundances (CPUF) and
metabarcoding taxonomic assignments and back-estimated
abundances (RPUF).
CPUF
RPUF
Loc 1 Loc 2 Loc 3
Lowest possible taxonomic level
Loc 1 −0.123 −0.013
Loc 2 −0.111 0.209*
Loc 3 −0.053 0.233*
Family level
Loc 1 −0.130 −0.073
Loc 2 −0.123 0.205*
Loc 3 −0.087 0.220*
*Significant difference (P ≤0.05) in community composition
between two locations.
Article e02284; page 8 FRANCES C. RATCLIFFE ET AL. Ecological Applications
Vol. 31, No. 3

CONCLUSIONS
We have shown that using a single marker (12S), equal
amounts of tissue per sample and estimation of number
of individuals from RRA, metabarcoding can provide
quantitative abundance estimates for the calculation of
alpha and beta diversities. This method could be applied
to bulk samples from different terrestrial and marine
habitats to improve abundance estimates. Specifically,
we recommend the use of markers with highly conserved
binding sites and using a small, equally sized pieces of
tissue from each specimen to minimize biases and han-
dling steps. This provides a rapid, community level
assessment method, that could be used to further under-
stand responses to disturbance and inter-annual or sea-
sonal variability and monitor biodiversity in a changing
global climate.
ACKNOWLEDGMENTS
We are very grateful to those who assisted with sampling:
Helen McCormick, Ross O’Neill, Michael Sheridan, Sarah
Albuixech-Marti, Katie Costello, and crew of the R.V. Celtic
Voyager. This work has been funded by the European Regional
Development fund through the Ireland Wales Co-operation
Programme 2014–2020 (BlueFish project). Ethics approval:
sampling has been conducted following Home Office regula-
tions and approved by Swansea University Ethics Committees
under approval No. 181019/1996. S. Consuegra, F. C. Ratcliffe,
and C. G. de Leaniz conceived the ideas; F. C. Ratcliffe
designed the methodology with help of D. Rodriguez-Barreto;
F. C. Rarcliffe and T. M. U. Webster carried out the laboratory
work; F. C. Ratcliffe performed bioinformatic analyses with T.
M. U. Webster and R. O’Rorke; R. O’Rorke built the species
database; F. C. Rarcliffe led the writing of the manuscript with
S. Consuegra and all authors contributed critically to the drafts.
LITERATURE CITED
Acevedo, S., O. Dwane, and J. M. Fives. 2002. The community
structure of larval fish populations in an area of the Celtic
Sea in 1998. Journal of the Marine Biological Association of
the United Kingdom 82:641–648.
Alberdi, A., O. Aizpurua, M. T. P. Gilbert, and K. Bohmann.
2018. Scrutinizing key steps for reliable metabarcoding of
environmental samples. Methods in Ecology and Evolution
9:134–147.
Asch, R. G. 2015. Climate change and decadal shifts in the phe-
nology of larval fishes in the California Current ecosystem.
Proceedings of the National Academy of Sciences USA 112:
E4065–E4074.
Becker, R. A., N. G. Sales, G. M. Santos, G. B. Santos, and D.
C. Carvalho. 2015. DNA barcoding and morphological identi-
fication of neotropical ichthyoplankton from the Upper Paran´
a
and S˜
ao Francisco. Journal of Fish Biology 87:159–168.
Bohmann, K., A. Evans, M. T. P. Gilbert, G. R. Carvalho, S.
Creer, M. Knapp, D. W. Yu, and M. de Bruyn. 2014. Environ-
mental DNA for wildlife biology and biodiversity monitor-
ing. Trends in Ecology and Evolution 29:358–367.
Bolyen, E., et al. 2019. Reproducible, interactive, scalable and
extensible microbiome data science using QIIME 2. Nature
Biotechnology 37:852–857.
Borja, A., M. Elliott, M. C. Uyarra, J. Carstensen, and M.
Mea. 2017. Editorial: Bridging the gap between policy and
science in assessing the health status of marine ecosystems.
Frontiers in Marine Science 4:32. https://doi.org/10.3389/fma
rs.2017.00032
Boyer, F., C. Mercier, A. Bonin, Y. Le Bras, P. Taberlet, and E.
Coissac. 2016. obitools: A unix-inspired software package for
DNA metabarcoding. Molecular Ecology Resources
16:176–182.
Brechon, A. L., S. H. Coombs, D. W. Sims, and A. M. Griffiths.
2013. Development of a rapid genetic technique for the iden-
tification of clupeid larvae in the Western English Channel
and investigation of mislabelling in processed fish products.
ICES Journal of Marine Science 70:399–407.
Callahan, B. J., P. J. McMurdie, and S. P. Holmes. 2017. Exact
sequence variants should replace operational taxonomic units
in marker-gene data analysis. ISME Journal 11:2639–2643.
Canfield, T. J., and J. R. Jones. 1996. Zooplankton abundance,
biomass, and size-distribution in selected midwestern water-
bodies and relation with trophic state. Journal of Freshwater
Ecology 11:171–181.
Clarke, K. R. 1993. Non-parametric multivariate analyses of
changes in community structure. Australian Journal of Ecol-
ogy 18:117–143.
Clarke, K. R., and R. N. Gorley. 2015. Getting started with
PRIMER v7. PRIMER-E, Plymouth Marine Laboratory,
Plymouth, UK.
Collins, R. A., J. Bakker, O. S. Wangensteen, A. Z. Soto, L.
Corrigan, D. W. Sims, M. J. Genner, and S. Mariani. 2019.
Non-specific amplification compromises environmental DNA
metabarcoding with COI. Methods in Ecology and Evolution
10:1985–2001.
Deagle, B. E., S. N. Jarman, E. Coissac, F. Pompanon, and P.
Taberlet. 2014. DNA metabarcoding and the cytochrome c
oxidase subunit I marker: not a perfect match. Biology Let-
ters 10:2–5.
Deagle, B. E., A. C. Thomas, J. C. McInnes, L. J. Clarke, E. J.
Vesterinen, E. L. Clare, T. R. Kartzinel, and J. P. Eveson.
2019. Counting with DNA in metabarcoding studies: How
should we convert sequence reads to dietary data? Molecular
Ecology 28:391–406.
Dopheide, A., et al. 2019. Estimating the biodiversity of ter-
restrial invertebrates on a forested island using DNA bar-
codes and metabarcoding data. Ecological Applications 29:
e01877.
Edgar, R. 2010. Usearch. Lawrence Berkeley National Lab
(LBNL), Berkeley, California, USA.
Elbrecht, V., B. Peinert, and F. Leese. 2017. Sorting things out:
assessing effects of unequal specimen biomass on DNA
metabarcoding. Ecology and Evolution 7:6918–6926.
Ellis, J. R., S. P. Milligan, L. Readdy, N. Taylor, and M. J.
Brown. 2012. Spawning and nursery grounds of selected fish
species in UK waters. https://www.cefas.co.uk/publications/
techrep/TechRep147.pdf
Findley, K., et al. 2013. Topographic diversity of fungal and
bacterial communities in human skin. Nature 498:367–370.
Hall, T., I. Biosciences, and C. Carlsbad. 2011. BioEdit: An
important software for molecular biology. GERF Bulletin of
Biosciences 2:60–61.
Hernandez, F. J., L. Carassou, W. M. Graham, and S. P. Powers.
2013. Evaluation of the taxonomic sufficiency approach for
ichthyoplankton community analysis. Marine Ecology Pro-
gress Series 491:77–90.
Huson, D. H., A. F. Auch, J. Qi, and S. C. Schuster. 2007.
MEGAN analysis of metagenomic data. Genome Research
17:377–386.
Kimmerling, N., et al. 2018. Quantitative species-level ecology
of reef fish larvae via metabarcoding. Nature Ecology & Evo-
lution 2:306–316.
April 2021 QUANTITATIVE FISH LARVAE METABARCODING Article e02284; page 9

Krehenwinkel, H., M. Wolf, J. Y. Lim, A. J. Rominger, W. B.
Simison, and R. G. Gillespie. 2017. Estimating and mitigating
amplification bias in qualitative and quantitative arthropod
metabarcoding. Scientific Reports 7:17668. https://doi.org/10.
1038/s41598-017-17333-x
Lamb, P. D., E. Hunter, J. K. Pinnegar, S. Creer, R. G. Davies,
and M. I. Taylor. 2019. How quantitative is metabarcoding: a
meta-analytical approach. Molecular Ecology 28:420–430.
Legrand, T., A. Di Franco, E. Ser-Giacomi, A. Calo, and V.
Rossi. 2019. A multidisciplinary analytical framework to
delineate spawning areas and quantify larval dispersal in
coastal fish. Marine Environmental Research 151:104761.
Miya, M., et al. 2015. MiFish, a set of universal PCR primers
for metabarcoding environmental DNA from fishes: detec-
tion of more than 230 subtropical marine species. Royal Soci-
ety Open Science 2:150088.
Morgulis, A., G. Coulouris, Y. Raytselis, T. L. Madden, R.
Agarwala, and A. A. Sch¨
affer. 2008. Database indexing for
production MegaBLAST searches. Bioinformatics
24:1757–1764.
Nobile, A. B., D. Freitas-Souza, F. J. Ruiz-Ruano, M. L. M. O.
Nobile, G. O. Costa, F. P. de Lima, J. P. M. Camacho, F. For-
esti, and C. Oliveira. 2019. DNA metabarcoding of neotropi-
cal ichthyoplankton: Enabling high accuracy with lower cost.
Metabarcoding and Metagenomics 3:69–76.
Pi˜
nol, J., G. Mir, P. Gomez-Polo, and N. Agustı´. 2015. Universal
and blocking primer mismatches limit the use of high-
throughput DNA sequencing for the quantitative metabar-
coding of arthropods. Molecular Ecology Resources
15:819–830.
Pi˜
nol, J., M. A. Senar, and W. O. C. Symondson. 2019. The
choice of universal primers and the characteristics of the spe-
cies mixture determine when DNA metabarcoding can be
quantitative. Molecular Ecology 28:407–419.
Porter, T. M., and M. Hajibabaei. 2018. Scaling up: A guide to
high-throughput genomic approaches for biodiversity analy-
sis. Molecular Ecology 27:313–338.
R Core Team. 2020. R: A language and environment for statisti-
cal computing. R Foundation for Statistical Computing,
Vienna, Austria. https://www.R-project.org/
Radchuk, V., C. Turlure, and N. Schtickzelle. 2013. Each life
stage matters: the importance of assessing the response to cli-
mate change over the complete life cycle in butterflies. Jour-
nal of Animal Ecology 82:275–285.
Riaz, T., W. Shehzad, and A. Viari. 2011a. ecoPrimers: infer-
ence of new DNA barcode markers from whole genome
sequence analysis. Nucleic Acids Research 39:1–11.
Riaz, T., W. Shehzad, A. Viari, F. Pompanon, P. Taberlet, and
E. Coissac. 2011b. EcoPrimers: Inference of new DNA bar-
code markers from whole genome sequence analysis. Nucleic
Acids Research 39:1–11.
Richardson, R. T., C. Lin, J. O. Quijia, N. S. Riusech, K. Good-
ell, and R. M. Johnson. 2015. Rank-based characterization of
pollen assemblages collected by honey bees using a multi-
locus metabarcoding approach. Applications in Plant
Sciences 3:1500043.
Richardson, A. J., A. W. Walne, A. W. G. John, T. D. Jonas, J.
A. Lindley, D. W. Sims, D. Stevens, and M. Witt. 2006. Using
continuous plankton recorder data. Progress in Oceanogra-
phy 68:27–74.
Robinson, C. V., T. M. U. Webster, J. Cable, J. James, and S.
Consuegra. 2018. Simultaneous detection of invasive signal
crayfish, endangered white-clawed crayfish and the crayfish
plague pathogen using environmental DNA. Biological Con-
servation 222:241–252.
Rodriguez, J. M., F. Alemany, and A. Garcia. 2017. A guide to
the eggs and larvae of 100 common Western Mediterranean
Sea bony fish species. https://www.agris.fao.org
Rose, K. A. 2000. Why are quantitative relationships between
environmental quality and fish populations so elusive? Eco-
logical Applications 10:367–385.
Russell, F. S. 1976. The eggs and planktonic stages of British
marine fishes. Academic Press, London, UK.
Schloss, P. D., et al. 2009. Introducing mothur: Open-source,
platform-independent, community-supported software for
describing and comparing microbial communities. Applied
and Environmental Microbiology 75:7537–7541.
Schnell, I. B., K. Bohmann, and M. T. P. Gilbert. 2015. Tag
jumps illuminated—reducing sequence-to-sample misidentifi-
cations in metabarcoding studies. Molecular Ecology
Resources 15:1289–1303.
Sigut, M., M. Kostov´
cik, J. Hulcr, P. Drozd, and J. Hrˇ
cek. 2017.
Performance of DNA metabarcoding, standard barcoding,
and morphological approach in the identification of host–-
parasitoid interactions. PLoS ONE 12:e0187803–1.
Taberlet, P., E. Coissac, F. Pompanon, C. Brochmann, and E.
Willerslev. 2012. Towards next-generation biodiversity assess-
ment using DNA metabarcoding. Molecular Ecology
21:2045–2050.
Tang, M., et al. 2015. High-throughput monitoring of wild bee
diversity and abundance via mitogenomics. Methods in Ecol-
ogy and Evolution 6:1034–1043.
Thomas, A. C., B. E. Deagle, J. P. Eveson, C. H. Harsch, and A.
W. Trites. 2016. Quantitative DNA metabarcoding: improved
estimates of species proportional biomass using correction
factors derived from control material. Molecular Ecology
Resources 16:714–726.
Thomsen, P. F., P. R. Møller, E. E. Sigsgaard, S. W. Knudsen,
O. A. Jørgensen, and E. Willerslev. 2016. Environmental
DNA from seawater samples correlate with trawl catches of
subarctic, deepwater fishes. PLoS ONE 11:e0165252.
Ward, R. D., T. S. Zemlak, B. H. Innes, P. R. Last, and P. D. N.
Hebert. 2005. DNA barcoding Australia’s fish species. Philo-
sophical Transactions of the Royal Society B 360:1847–1857.
Wiesner, R. J., J. C. R ¨
uegg, and I. Morano. 1992. Counting tar-
get molecules by exponential polymerase chain reaction: copy
number of mitochondrial DNA in rat tissues. Biochemical
and Biophysical Research Communications 183:553–559.
SUPPORTING INFORMATION
Additional supporting information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/eap.2284/full
DATA AVAILABILITY
Sequences from metabarcoding have been deposited in the NCBI under accession reference BioProject PRJNA576002. Sanger
sequences for the reference collection have been deposited in GenBank under accession numbers MN539918–MN539945 (CO-I)
and MN539946–MN539976 (12S).
Article e02284; page 10 FRANCES C. RATCLIFFE ET AL. Ecological Applications
Vol. 31, No. 3